• OveRview
  • What is R, brief history
• How to get setup and use R
• Fundamentals of R
  • Basic syntax
  • Data classes and structures [Main Section]
  • Reading data in
  • Installing/loading packages
• Staying up-to-date in R/further reading

R is:

• A statistical programming, computation, and graphics system,
• Its own language (code-based programming/commands),
• A cross platform system (windows, mac, unix-like),
• Freely available online,
• A modular system

R is an environment where you can read data in, carry out data manipulation, analyse your data, and prepare your results for others to see.

R is an environment where you can:

• read (pull) data in (from a variety of sources),
• carry out (very complex) data manipulation,
• analyse your data (via a broad and rapidly increasing suite of statistical models), and
• prepare (reproducibly) your results (figures, tables, reports, papers) for others to see (or interact with).

• R first appeared in 1993
• Originally written by Ross Ihaka and Robert Gentleman (University of Auckland, NZ)
• Influenced by the existing languages S and Scheme
• Mid 1997, ‘R core team’ established - maintain the base code
• R CRAN (Comprehensive R Archive Network)
  • A collection of sites (mirrored daily around the world) that manages the distribution of R and its packages
• R V1.0 released 2000-02-29 (stable)

• Over 10,000 R ‘add-on’ packages (libraries) on CRAN
  • +bioconductor, +github etc
• RStudio initially released in 2011
  • integrated development environment
• Annual conference
  • useR! (2018 Brisbane, 2019 Toulouse)
• R-Ladies

• 200? enter Hadley Wickham
  • ggplot package
  • plyr package
• Chief Scientist at RStudio
  • Kiwi
  • Developed many packages
  • Written many books (R for Data Science, Advanced R)
    • Freely available online

• Tidyverse
  • controversy
  • base vs tidyverse

“The tidyverse is an opinionated collection of R packages designed for data science. All packages share an underlying design philosophy, grammar, and data structures.”

R is not a:

• data collection tool
• (permanent) data storage system

• There are many ways to use R
• There are many ways to do most operations R
• This presentation focuses on the collective ‘best practice’ advice of our team of daily R users
• R is rapidly evolving
• Online help/guides are often out-of-date

• R is an interactive language
  • Type a command, submit, immediate response (line-by-line)
• Typically, commands saved in a plain text file (.R)
• Commands can be executed in bulk
  • Large scale analysis/data manipulation
  • ‘one-click’ reports (like these slides)

• R is a cAsE sensitive language
• R is largely forgiving with where white space is (or isn’t!)
  • including line breaks, bracket placement etc

• Download R for your OS
  • https://cran.r-project.org/
• You will need admin rights
• Download RStudio (optional, well, not really)
  • https://www.rstudio.com/products/rstudio/download/

• free and open-source integrated development environment for R
• Screen divided into quarters, typically
  • One for your code (commands)
  • One for your console (commands executed/results shown)
  • One for your environment management
  • One for your plots/help

• Customisable and quite diverse
• Ready access to:
  • Terminal, code repository, package management, file management
• Other benefits including
  • Code highlighting (full-featured)
  • Customisable settings (load/save session)
  • Tab completion

• Assign ‘<-’
  • less than sign followed by dash
  • puts the right hand side into the left hand side

a <- 3

• a will now be 3 (which we’ll see shortly)
• You can use ‘=’ in place of ‘<-’
• Typically we only use ‘=’ within brackets
  • It’s complicated and related to environments (scoping)

• basic math

3+3

## [1] 6

3*3

## [1] 9

a <- 3
a + a

## [1] 6

• Round brackets (something) | (x, y)
  • multiple things inside, separated by a comma
  • typically used to group/provide arguments for a function

sum(1, 2, 3, 4, NA, 5, na.rm = T)

## [1] 15

• Square brackets [something] | [ ] [,] [[ ]]
  • multiple things inside, separated by a comma
  • typically used for indices when navigating within a structure

c("one", "two", "three")[2]

## [1] "two"

• Curly brackets {something}
  • groups multiple lines of code
  • typically used when creating functions

## [1] "matt is awesome"

• Colon : used to create a sequence
  • 1:4 produces 1 2 3 4
  • double colon to access functions in libraries that aren’t loaded
• Hashtag for comments # For this next bit of code online somewhere
  • Code won’t be executed
• Question mark ?something for help
  • Double question mark to search for commands

• The letters T and F are ‘reserved’ - logicals
  • indicate TRUE and FALSE respectively, don’t overwrite them
  • used regularly within arguments
• Working directory
  • getwd() and setwd()
  • best practise is using relative directory paths!

• Each data point (value within a cell) will have a class.
• There are many classes, most (and the common ones) are nice and clear.
• Classes have, somewhat of, a hierarchy
  • character is King/Queen amongst many
• Classes are closely related to Structures (the next topic).
• Functions will respond differently based on the class of data you feed them.
• Sometimes referred to as mode

• character is the highest-level ‘catch-all’ class
• Free text fields will be of class character
• Any* data point that has something other than numbers in it will default to being of class character

*data points with non-nuermic (alpha) characters can be made into other classes (more to come)

x <- "Teach me R"
x

## [1] "Teach me R"

class(x)

## [1] "character"

• Note the quote marks

x <- Teach me R

## Error: <text>:1:12: unexpected symbol
## 1: x <- Teach me
##                ^

• Note the lack of quote marks

x <- "Teach me R"
y <- "Teach me now"
x + y

## Error in x + y: non-numeric argument to binary operator

x <- "3"
class(x)

## [1] "character"

• Interesting?

• There are three classes that can all be thought of as “numbers”
  • numeric
  • double (double precision floating point numbers - computer format)
  • integer (no decimals)

• R may convert between the three as it sees fit

• “it is perfectly feasible to use R successfully for years and not need to know the answer to this question” - stackoverflow

x <- "3"
x

## [1] "3"

class(x)

## [1] "character"

• Oh, quotes!

x <- 3
x

## [1] 3

class(x)

## [1] "numeric"

x + 2

## [1] 5

x <- as.integer(3.1)
x

## [1] 3

class(x)

## [1] "integer"

• Notice the decimal has dropped off?

x + 0.4

## [1] 3.4

class(x + 0.4)

## [1] "numeric"

• Notice the decimal has dropped off

• Class factor is used when there is a limited set of responses
  • e.g. likert scales, States, income brackets
• Generally have labels, associated with levels (the responses)
• Always have an underlying number linked to each label/level
• Can be ordered, but don’t have to be
  • Default behaviour is to handle alphabetically
• Useful for creating tables, coefficients in models etc

y <- c("Cat", "Other", "Dog", "Dog"); y

## [1] "Cat"   "Other" "Dog"   "Dog"

class(y)

## [1] "character"

y <- as.factor(c("Cat", "Other", "Dog", "Dog")); y

## [1] Cat   Other Dog   Dog  
## Levels: Cat Dog Other

• note, the order of the levels

class(y); levels(y)

## [1] "factor"

## [1] "Cat"   "Dog"   "Other"

y <- y[-1]  # get rid of the cat
table(y)

## y
##   Cat   Dog Other 
##     0     2     1

y <- as.factor(y); levels(y)

## [1] "Cat"   "Dog"   "Other"

y <- factor(y); levels(y)

## [1] "Dog"   "Other"

• review ?factor
• as.factor() does not accept additional arguments
  • see levels defaults.
• relevel lets you set the ‘baseline’ level
  • very handy for modeling

class(tki_demo$intervention)

## [1] "factor"

table(tki_demo$intervention)

## 
## Placebo  Drug 1  Drug 2 
##      29      38      33

levels(tki_demo$intervention)      

## [1] "Placebo" "Drug 1"  "Drug 2"

levels(factor(as.character(tki_demo$intervention)))

## [1] "Drug 1"  "Drug 2"  "Placebo"

y <- tki_demo$intervention[1:5]
y           # retained class and levels

## [1] Drug 2  Drug 2  Drug 2  Placebo Drug 1 
## Levels: Placebo Drug 1 Drug 2

y <- y[-5]  # get rid of the 'Drug 1'
table(y)

## y
## Placebo  Drug 1  Drug 2 
##       1       0       3

y <- as.factor(y); levels(y)

## [1] "Placebo" "Drug 1"  "Drug 2"

y <- factor(y); levels(y)

## [1] "Placebo" "Drug 2"

• Class logical variables can only take two*, TRUE or FALSE
• Can usefully be abbreviated as T and F
• Generally used as a selection mechanism
  • records in/out
  • conditional (if) statements
• Work well with sum(), any(), all()
• Be aware of combined logicals!
  • F + F + NA = ?
  • all(c(T, NA, T)) = ?

y <- c(TRUE, FALSE, TRUE, FALSE, FALSE)
class(y)

## [1] "logical"

table(y)

## y
## FALSE  TRUE 
##     3     2

F + F + NA

## [1] NA

T + NA + T

## [1] NA

F + F + F

## [1] 0

T + T + T

## [1] 3

Oh!

table(y)

## y
## FALSE  TRUE 
##     3     2

sum(y)

## [1] 2

all(c(T, NA, T))

## [1] NA

all(c(F, NA, F))

## [1] FALSE

any(c(T, NA, T))

## [1] TRUE

any(c(F, NA, F))

## [1] NA

all(c(T, F, T))

## [1] FALSE

• Class date follows relatively strict formatting
• Can also include time
  • via ‘associated’ classes POSIXct, POSIXlt
  • be aware of timezones
• Relatively nice syntax for coverting strings to dates
• Has many helper funtions to aid date math
• Lubridate!
• Go slow!

• ?strptime can be your friend
  • Date-time Conversion Functions to and from Character

%d Day of the month as decimal number (01–31).

%e Day of the month as decimal number (1–31), …

%H Hours as decimal number (00–23). …

%y Year without century (00–99). On input, values 00 to 68 are prefixed by 20 and 69 to 99 by 19…

%Y Year with century. Note that whereas there was no zero in the original Gregorian calendar…

• Note the structure

Sys.Date()

## [1] "2019-07-26"

class(as.Date("2019-01-31"))

## [1] "Date"

class(as.Date("31-01-2019"))

## [1] "Date"

class(as.Date("31.01.2019"))

## Error in charToDate(x): character string is not in a standard unambiguous format

class(as.Date("31.01.2019", format = "%d.%m.%Y"))

## [1] "Date"

• library(lubridate)

library(lubridate); class(dmy("31.01.2019"))

## [1] "Date"

• complex (powerful?) new classes
  • duration, interval, period

round(interval(ymd("1983-10-16"), Sys.Date()) / years(1), 2)

## [1] 35.78

as.duration(interval(ymd("1983-10-16"), Sys.Date()))

## [1] "1128988800s (~35.78 years)"

as.period(interval(ymd("1983-10-16"), Sys.Date()))

## [1] "35y 9m 10d 0H 0M 0S"

• Structure is our term, it is not an R term
• Sometimes referred to as type
• Some structures are a class within R
• But not all

Single element (scalar typically) - most basic building block

i <- 1

• at least two elements
• a vector is not a class
• a vector HAS a class, and it can have only one! (remember the hierarchy)
• the power of R, is vectorised processing

x <- c(3,4,5,6,8) # c stands for combine/concatenate
x

## [1] 3 4 5 6 8

x + 2

## [1]  5  6  7  8 10

length(x)

## [1] 5

sum(x)

## [1] 26

y <- c("Cat", "Dog", "Other", "Dog")
length(y)

## [1] 4

sum(y)

## Error in sum(y): invalid 'type' (character) of argument

• Likely the main* structure you will work with in R
• Best ‘pictured’ as a table (two-dimensional), with rows and columns
• Best ‘thought of’ as:
  • a number of | vertical | vectors, side-by-side, bound together, all of equal length
• All elements of a column will be off the same class!
• When reading in a csv, a plain text file, (a worksheet of) an excel file
  • it will generally become available to you as a data.frame
• standby for tibbles

• A tibble is functionally very similar to a data.frame
  • Part of the tidyverse
  • Has a “handy” reduced display format

We’ll come back to this

class(tki_demo)

## [1] "data.frame"

dim(tki_demo)

## [1] 100   8

head(tki_demo) # can you guess what tail does?

##   id        dob  male smoker intervention      day1      day2     day3
## 1  1 2004-12-08  TRUE  FALSE       Drug 2  3.787324 19.379647 29.63681
## 2  2 2007-06-14 FALSE   TRUE       Drug 2  1.200292 28.770240       NA
## 3  3 2003-01-05  TRUE  FALSE       Drug 2  6.321257 22.820082 39.02544
## 4  4 2002-09-14 FALSE  FALSE      Placebo -1.302337  4.610366  9.40575
## 5  5 2003-10-24 FALSE  FALSE       Drug 1  7.793055 19.879646 14.73297
## 6  6 2009-03-06  TRUE  FALSE       Drug 1 11.310929 12.648613       NA

names(tki_demo)

## [1] "id"           "dob"          "male"         "smoker"      
## [5] "intervention" "day1"         "day2"         "day3"

str(tki_demo)

## 'data.frame':    100 obs. of  8 variables:
##  $ id          : int  1 2 3 4 5 6 7 8 9 10 ...
##  $ dob         : Date, format: "2004-12-08" "2007-06-14" ...
##  $ male        : logi  TRUE FALSE TRUE FALSE FALSE TRUE ...
##  $ smoker      : logi  FALSE TRUE FALSE FALSE FALSE FALSE ...
##  $ intervention: Factor w/ 3 levels "Placebo","Drug 1",..: 3 3 3 1 2 2 3 1 1 1 ...
##  $ day1        : num  3.79 1.2 6.32 -1.3 7.79 ...
##  $ day2        : num  19.38 28.77 22.82 4.61 19.88 ...
##  $ day3        : num  29.64 NA 39.03 9.41 14.73 ...

summary(tki_demo[, 1:4])

##        id              dob                male           smoker       
##  Min.   :  1.00   Min.   :1899-02-25   Mode :logical   Mode :logical  
##  1st Qu.: 25.75   1st Qu.:2003-06-06   FALSE:65        FALSE:72       
##  Median : 50.50   Median :2005-07-12   TRUE :35        TRUE :28       
##  Mean   : 50.50   Mean   :2004-05-07                                  
##  3rd Qu.: 75.25   3rd Qu.:2007-04-02                                  
##  Max.   :100.00   Max.   :2009-04-13

head(tki_demo)

##   id        dob  male smoker intervention      day1      day2     day3
## 1  1 2004-12-08  TRUE  FALSE       Drug 2  3.787324 19.379647 29.63681
## 2  2 2007-06-14 FALSE   TRUE       Drug 2  1.200292 28.770240       NA
## 3  3 2003-01-05  TRUE  FALSE       Drug 2  6.321257 22.820082 39.02544
## 4  4 2002-09-14 FALSE  FALSE      Placebo -1.302337  4.610366  9.40575
## 5  5 2003-10-24 FALSE  FALSE       Drug 1  7.793055 19.879646 14.73297
## 6  6 2009-03-06  TRUE  FALSE       Drug 1 11.310929 12.648613       NA

tki_demo[1:2 , 1:3]

##   id        dob  male
## 1  1 2004-12-08  TRUE
## 2  2 2007-06-14 FALSE

tki_demo[1 , 1]

## [1] 1

tki_demo$day1[1:10]

##  [1]  3.787324  1.200292  6.321257 -1.302337  7.793055 11.310929  8.389122
##  [8]  9.046324  6.428155  3.388006

tki_demo$index <- 1:10
tki_demo$index <- 1:nrow(tki_demo)
head(tki_demo)

##   id        dob  male smoker intervention      day1      day2     day3
## 1  1 2004-12-08  TRUE  FALSE       Drug 2  3.787324 19.379647 29.63681
## 2  2 2007-06-14 FALSE   TRUE       Drug 2  1.200292 28.770240       NA
## 3  3 2003-01-05  TRUE  FALSE       Drug 2  6.321257 22.820082 39.02544
## 4  4 2002-09-14 FALSE  FALSE      Placebo -1.302337  4.610366  9.40575
## 5  5 2003-10-24 FALSE  FALSE       Drug 1  7.793055 19.879646 14.73297
## 6  6 2009-03-06  TRUE  FALSE       Drug 1 11.310929 12.648613       NA
##   index
## 1     1
## 2     2
## 3     3
## 4     4
## 5     5
## 6     6

• a tibble is the data.frame of the tidyverse
• broady the same as a data.frame, but less unruly (print)

load("../data/dat.RData"); head(tki_demo, 3)

## # A tibble: 3 x 8
##      id dob        male  smoker intervention  day1  day2  day3
##   <int> <date>     <lgl> <lgl>  <fct>        <dbl> <dbl> <dbl>
## 1     1 2004-12-08 TRUE  FALSE  Drug 2        3.79  19.4  29.6
## 2     2 2007-06-14 FALSE TRUE   Drug 2        1.20  28.8  NA  
## 3     3 2003-01-05 TRUE  FALSE  Drug 2        6.32  22.8  39.0

head(tki_demo, 2)

## # A tibble: 2 x 8
##      id dob        male  smoker intervention  day1  day2  day3
##   <int> <date>     <lgl> <lgl>  <fct>        <dbl> <dbl> <dbl>
## 1     1 2004-12-08 TRUE  FALSE  Drug 2        3.79  19.4  29.6
## 2     2 2007-06-14 FALSE TRUE   Drug 2        1.20  28.8  NA

• Given the dimensions by default
• Given the column classes by default
• Control printing (won’t blow out your console)

• Best ‘pictured’ as a 3D (deep) (filing cabinet like) structure
  • A vector where each element can have a different class/dimension
• Lists can be used to keep related things together rather than as separate objects
  • If you had a data.frame of data for each calendar
  • You can apply operations to call calendar years at the same time
• Lists can be good for collecting output (complex elements) from a loop
  • Model output from running a model on multiple imputed datasets

tki_list <- list(tki_demo,
                 tki_demo_complications)
class(tki_list)

## [1] "list"

head(tki_list[[1]], 2) # access the first element 

## # A tibble: 2 x 8
##      id dob        male  smoker intervention  day1  day2  day3
##   <int> <date>     <lgl> <lgl>  <fct>        <dbl> <dbl> <dbl>
## 1     1 2004-12-08 TRUE  FALSE  Drug 2        3.79  19.4  29.6
## 2     2 2007-06-14 FALSE TRUE   Drug 2        1.20  28.8  NA

head(tki_list[[2]], 2) # access the second element 

## # A tibble: 2 x 2
##      id complications
##   <int> <chr>        
## 1    12 Yoda speach  
## 2    22 Man flu

• why? efficiency!

lapply(tki_list, head, 2)

## [[1]]
## # A tibble: 2 x 8
##      id dob        male  smoker intervention  day1  day2  day3
##   <int> <date>     <lgl> <lgl>  <fct>        <dbl> <dbl> <dbl>
## 1     1 2004-12-08 TRUE  FALSE  Drug 2        3.79  19.4  29.6
## 2     2 2007-06-14 FALSE TRUE   Drug 2        1.20  28.8  NA  
## 
## [[2]]
## # A tibble: 2 x 2
##      id complications
##   <int> <chr>        
## 1    12 Yoda speach  
## 2    22 Man flu

Matrix (2-dimensions)

• A grid like structure (similar to data.frames)
• Entire strucutre is one class (all numeric or character etc)
• Can not use $ for referencing columns ([ , ] used)

Array (n-dimensions)

• Like a list, but must be all of the same class

beyond the scope of this level - computational efficiency

• The function you use depends on your file type
• Flat files (plain text files) (.csv, .txt) can be read in with base functions
  • read.csv() and read.table() are commonly use
• Other files require specific packages (next topic)
  • The readxl package reads in (.xlx, .xlsx) excel files via read_excel()
  • The foreign and haven packages read in SPSS, Stata, and SAS files

• header = T
  • Is the top row of your dataset column headers
• na.strings = c(“NA”, “missing”, “999”)
  • This will mean cells end up as true R based NA values, which is very handy
• stringsAsFactors = F
  • Can cause issues if not explicitly set as F (ID codes, free text responses)

dat <- read.csv(“/DIRECTORY/my_data_file.csv”, header = T, na.strings = c(“NOT CONTACTED”))

or

dat <- read.csv(“../data/my_data_file.csv”, header = T, na.strings = c(“NOT CONTACTED”))

• This will load the csv file into a data.frame called dat
• Ensure path is valid (shared drives)

• Excel
  • skip lines? data classes (dates?)?
  • which worksheet(s?) did you want?
• SPSS
  • do you want category labels? or underlying stored values (think factors)?

• A single character in a column that isn’t 0-9 or . will mean that column is read in as character
• You can load data directly from the web
  • read.csv(“http://some.where.net/data/foo.csv”)
  • You can use readHTMLTable from the XML package to read tables off websites

• Packages can be thought of as ‘bundles of functions’
  • They can contain datasets and other things too
• They extend the functionality of R
  • New graphing abilities, modeling methods, faster data manipulation
• It’s very common for scripts to start by calling many packages
• Require semi-regular updating (easy)

• Installed via: install.packages(“ggplot2”)
  • note the quote marks
  • by default, this will also install that packages dependencies
• Packages loaded for use via: library(ggplot)
  • note the lack of quote marks
• Updated via: updated.packages()
  • considered using ‘ask = T’

• Some packages are installed direct from the developer
  • Tnstalled via devtools::install_github()
  • That is, using the install_github() function within the devtools package
• Our package is installed this way
  • devtools::install_github(“TelethonKids/biometrics”, build_vignettes = TRUE)

• Running .libPaths() will show where (on your computer) your packages are stored
• Running citation(“ggplot2”) will tell you how to cite the package
• Beware of clashes
  • packages overiding base functions, or two packages with the same function
  • not overly common, and you will get warned
• Help file will tell you what package a function resides in
  • ?geom_path starts (top left) with geom_path {ggplot2}

• Save all scripts (.R, .Rmd) so you can *always regenerate your results from your original data files
• Saving your workspace is commonly done
  • This will include *everything within your current session (/environment)
  • All datasets, model output, custom functions, settings changes - but not loaded libraries
• Saved via: save.image(“my_session.Rdata”)
• Loaded via: load(“my_session.Rdata”)

• Save specific (individual) structures/object on their own (not commonly done)
  • saved via: saveRDS(dat, “my_data_frame.RDS”)
  • loaded via: readRDS(“my_data_frame.RDS”)
• Exporting data via write commands
  • export via: write.csv(dat, “my_csv_export.csv”)
  • commonly used arguments include row.names = F and na = ""

• R-bloggers https://www.r-bloggers.com/
• Stackoverflow https://stackoverflow.com/questions/tagged/r
• Twitter #rstats https://twitter.com/search?q=%23rstats
• UCLA’s educational resources https://stats.idre.ucla.edu/other/dae/
• Hadley resources
• Coding style guides

Many of these have been seen during the prior examples: - <- assign, puts the right hand side into the left hand side - ? followed by a command, to search for help on a command - a:b used to generate a series of integers, from a to b - - function( , , …) arguments to a function are seperated by commas - c() concatenate, used to create a vector - c() concatenate, used to create a vector - [ ] [,] [[ ]] used to ‘extract’/interact with componets of a
- - -